nature genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Exome sequencing identifies rare damaging variants in ATP8B4 and ABCA1 as risk factors for Alzheimer’s disease Received: 31 July 2021 A list of authors and their affiliations appears at the end of the paper Accepted: 19 September 2022 Alzheimer’s disease (AD), the leading cause of dementia, has an estimated Published online: 21 November 2022 heritability of approximately 70%1. The genetic component of AD has been mainly assessed using genome-wide association studies, which do Check for updates not capture the risk contributed by rare variants2. Here, we compared the gene-based burden of rare damaging variants in exome sequencing data from 32,558 individuals—16,036 AD cases and 16,522 controls. Next to variants in TREM2, SORL1 and ABCA7, we observed a significant association of rare, predicted damaging variants in ATP8B4 and ABCA1 with AD risk, and a suggestive signal in ADAM10. Additionally, the rare-variant burden in RIN3, CLU, ZCWPW1 and ACE highlighted these genes as potential drivers of respective AD-genome-wide association study loci. Variants associated with the strongest effect on AD risk, in particular loss-of-function variants, are enriched in early-onset AD cases. Our results provide additional evidence for a major role for amyloid-β precursor protein processing, amyloid-β aggregation, lipid metabolism and microglial function in AD. Beyond autosomal-dominant early-onset AD (<1% of all AD cases, After sample QC, we first compared gene-based rare-variant bur- onset at ≤65 years), the common complex form of AD has an estimated dens between 12,652 AD cases, consisting of 4,060 early-onset AD cases heritability of approximately 70%1. Using genome-wide association (EOAD, age at onset ≤65 years) and 8,592 late-onset AD cases (LOAD, age studies (GWAS), 75 mostly common genetic risk factors/loci have been at onset >65 years) and 8,693 controls (stage 1 analysis; Supplementary associated with AD risk in populations with European ancestry; how- Table 3). We detected 7,543,193 variants after sample and variant QC ever, individually these common variants have low effect sizes2. Using and annotated LOF variants with LOFTEE and missense variants with the DNA sequencing strategies, rare (allele frequency <1%) damaging Rare Exome Variant Ensemble Learner (REVEL) score and selected vari- missense or loss-of-function (LOF) variants in the TREM2, SORL1 and ants with a minor allele frequency (MAF) < 1% (Supplementary Table 4). ABCA7 genes were identified to also contribute to the heritability of We defined 4 deleteriousness thresholds by incrementally including AD, with substantially higher effect sizes than individual GWAS hits3–8. variants with lower levels of predicted deleteriousness: LOF (n = 57,543), To detect additional genes for which rare variants are associated with LOF + REVEL ≥ 75 (n = 111,755), LOF + REVEL ≥ 50 (n = 211,665) and AD risk, it is necessary to compare genetic sequencing data from LOF + REVEL ≥ 25 (n = 409,733), respectively. Of the 19,822 autosomal thousands of AD cases and controls. In a large collaborative effort, protein-coding genes, we analyzed the 13,222 genes that had a cumula- we harmonized sequencing data of studies from Europe and the USA tive minor allele count (cMAC) ≥ 10 for the lowest deleterious threshold and applied a multistage gene burden analysis (Fig. 1a) (for sample LOF + REVEL ≥ 25 (Methods); 9,168 genes for the LOF + REVEL ≥ 50 descriptions, see Supplementary Table1 and Extended Data Figs. 1 threshold, 5,694 for the LOF + REVEL ≥ 75 threshold and 3,120 genes and 2). We observed site-specific technical biases, since data were for the LOF-only threshold (Fig. 1b). For these different deleteriousness generated at multiple centers, using heterogeneous methods (Sup- thresholds, this analysis has an estimated power of 41, 22, 11 and 4%, plementary Table 2). To account for these batch effects, we designed respectively to attain a signal with P < 1 × 10−6 in stage 1, assuming that and applied comprehensive quality control (QC) procedures (Meth- for a gene, the differential variant burden between cases and controls ods and Supplementary Tables 3–5). is associated with an odds ratio (OR) of 10.0 in EOAD and 3.33 in LOAD e-mail: h.holstege@amsterdamumc.nl; m.hulsman1@amsterdamumc.nl; gaelnicolas@hotmail.com; jean-charles.lambert@pasteur-lille.fr Nature Genetics | Volume 54 | December 2022 | 1786–1794 1786 Letter https://doi.org/10.1038/s41588-022-01208-7 a b Cumulative carrier frequency of selected variants for protein-coding genes 20,000 Total number of autosomal protein-coding genes (b37): n = 19,822 cumulative carrier frequency of prioritized variants Discovery Confirmation 17,500 Potential hits Number of genes that exceed a certain Not analyzed LOF Stage 1 Stage 2 in this study Significant LOF + REVEL ≥ 75 12,652 AD cases, 8,693 controls 3,384 AD cases, 7,829 controls MTC: FDR < 0.1 genes (including exome extracts) 15,000 LOF + REVEL ≥ 50 LOF + REVEL ≥ 25 Meta-analysis Stage 1 + stage 2 n = 13,222 genes analyzed MTC: Holm–Bonferroni, P < 0.05 12,500 Variant characterization Mega-sample 10,000 15,808 AD cases, 16,097 controls (stages 1+ 2, controlled for identity by decent) 7,500 GWAS gene burden analysis and 5,000 variant characterization Mega-sample count ≤ 10 Damaging without exome extracts 2,500 TR EM SOR 2 allele 14,977 AD cases, 14,750 controls L1 MTC: FDR < 0.05 ABCA7 0 10–4 10–3 10–2 Cumulative carrier frequency of prioritized variants per gene c APP, PSEN1, Power to attain Quantile–quantile plot for stage 1 PSEN2 P < 1 × 10−6 for stage 1 (21,345 samples) 40.0:10.0 Family studies Not 1,000 samp 17.5 LOF SORL1 exp LOF + REVEL ≥ 75 Previously e TREM2 explored: cted LOF + REVEL ≥ 50 15.0 20.0:5.0 l e s: LOF + REVEL ≥ 25 SORL1 5,000 SORL1 Observed P values (–log10) 80% Damaging SORL1, sam 12.5 Protective po TREM2, OR (EOAD:LOAD) we ple r s: 10.0:3.33 ABCA7 80 10.0 40 % APOE 80 60 po % Not analyzed % we in this study 20 r 7.5 % % ABCA7 Damaging allele count ≤ 50 Damaging allele count ≤ 25 ADAM 10 5.0:2.5 Damaging allele count ≤ 10 ATP8B4 ABCA1 ABCA7 5.0 SORL1 GWAS studies ATP8B4 3.5:2.0 ADAM10 ATP8B4 2.5 2.5:1.5 0.1 FDR < Inflation: 1.046 0 1.5:1.3 0 1 2 3 4 10–4 10–3 10–2 Expected P values (–log10) Cumulative carrier frequency of prioritized variants per gene Fig. 1 | Study setup and power. a, Schematic of the study setup. The AD from stage 1). Cumulative carrier frequency and estimated effect size ranges are association of genes identified in stage 1 was confirmed in stage 2 and indicated for common variants identified to associate with AD by GWAS (green), significance was determined by meta-analysis. Variant characteristics were rare-variant burdens in SORL1, TREM2 and ABCA7 identified using sequencing investigated in a merged mega-sample rather than the meta-sample, allowing studies3–7 (grey/blue), and for rare variants observed in autosomal dominant more accurate variant effect size estimates for variant categories/age-at-onset AD (magenta). Common variants with high effect sizes (red) are not expected to bins. The mega-sample (without exome extracts) was also used for the GWAS exist. Genes with cMAC < 10 were not analyzed (pink). Power calculations show gene burden analysis. MTC, multiple testing correction. b, Top, number of genes that aggregating more cases and controls might allow for the identification of (y axis) with at least a certain cumulative carrier frequency of prioritized variants rare-variants that have a large effect on AD but for which only few carriers are (x axis), prioritized according to different deleteriousness thresholds. White observed, or for variants that have a modest/average effect on AD, for which box, genes with a cMAC ≥ 10 (cumulative minor allele count of ≥10 prioritized many carriers are observed (power calculations shown in Supplementary alleles identified across the 12,652 cases and 8,693 controls in the stage 1 sample) Table 6). c, Quantile–quantile plot of P values determined in the stage 1 discovery were considered to have sufficient carrier frequency to allow burden analysis. analysis based on an ordinal logistic burden test. For each of 13,222 genes, The SORL1, TREM2 and ABCA7 genes are indicated, revealing that carriers of rare we tested the burden of variants adhering to four variant deleteriousness damaging variants in these genes are relatively common, allowing identification thresholds, conditional on having a cMAC ≥ 10 (n = 31,204 tests). Threshold for in smaller sample sizes3–7. Bottom, power analysis for stage 1, to attain a multiple testing correction: FDR < 0.1, P value inflation, 1.046. Gene names in P < 1 × 10−6, at the same scale as the top figure. For comparison, we indicate 80% black indicate the deleteriousness threshold of the most significant burden test power thresholds for sample sizes of 1,000 and 5,000 individuals (subsampled in that gene. (Supplementary Table 6). Therefore, this analysis has only the power all the burden tests performed, 13 tests, covering 6 genes, indicated a to discover genes for which either the differential variant burden is differential rare-variant burden between AD cases and controls (false associated with a large effect size, and/or genes for which large num- discovery rate (FDR) < 0.1): SORL1, TREM2, ABCA7, ATP8B4, ADAM10 bers of damaging variant carriers are observed (Fig. 1b). Using ordinal and ABCA1 (Table 1)). logistic regression, 31,204 burden tests were performed across 13,222 To confirm these signals, we applied an analysis model consistent genes in stage 1 (single genes were tested with up to 4 thresholds). with stage 1 to an independent stage 2 dataset, which after QC, con- Statistical inflation of test results was negligible (𝝀 = 1.046; Fig. 1c). Of sisted of 3,384 cases and 7,829 controls (Supplementary Table 3–5) Nature Genetics | Volume 54 | December 2022 | 1786–1794 1787 Letter https://doi.org/10.1038/s41588-022-01208-7 Table 1 | Stages 1 and 2 and meta-analysis AD association statistics Stage 1 (n = 21,345) Stage 2 (n = 11,213) Meta-analysis (n = 32,558) Gene Variant P FDR No. variants/ Case/control Pa No. variants/ Case/control P Holm– Case/control P hetero deleteriousness no. carriers OR (95% CI) no. carriers OR (95% CI) Bonferroni OR (95% CI) genous threshold SORL1 LOF + REVEL ≥ 25 4.8 × 10−6 0.017 242/917 1.3 (1.1–1.5) 1.3 × 10−6 122/478 1.5 (1.2–1.9) 1.5 × 10−10 4.7 × 10−6 1.4 (1.2–1.5) 1.6 × 10−1 LOF + REVEL ≥ 50 4.0 × 10 −18 <0.0001 167/290 2.6 (2.0–3.2) 1.4 × 10 −9 79/137 2.4 (1.7–3.5) 8.1 × 10 −26 2.5 × 10 −21 2.5 (2.1–3.1) 9.8 × 10−1 LOF + REVEL ≥ 75 1.1 × 10 −14 <0.0001 96/164 3.3 (2.4–4.6) 5.2 × 10 −10 45/82 3.9 (2.3–6.6) 1.1 × 10 −22 3.4 × 10 −18 3.5 (2.7–4.6) 4.3 × 10−1 LOF 4.7 × 10−15 <0.0001 37/48 15.6 (3.7–37.3) 1.6 × 10−6 16/20 16.3 (3.8–35.0) 3.3 × 10−18 1.0 × 10−13 16.0 (9.5–27.0) 9.4 × 10−1 TREM2 LOF + REVEL ≥ 25 2.6 × 10−16 <0.0001 17/291 3.6 (2.9–4.6) 1.6 × 10−7 12/155 2.4 (1.6–3.4) 5.2 × 10−22 1.6 × 10−17 3.2 (2.6–3.9) 6.5 × 10−1 ABCA7 LOF + REVEL ≥ 25 9.5 × 10−8 0.001 265/959 1.4 (1.2–1.6) 9.8 × 10−8 170/502 1.6 (1.3–2.0) 4.1 × 10−13 1.3 × 10−8 1.4 (1.3–1.6) 6.5 × 10−2 LOF + REVEL ≥ 75 4.6 × 10 −6 0.017 93/297 1.6 (1.3–2.1) 4.8 × 10 −4 54/167 1.8 (1.3–2.6) 7.3 × 10 −9 2.3 × 10 −4 1.7 (1.4–2.1) 9.1 × 10−1 ATP8B4 LOF + REVEL ≥ 25 7.2 × 10 −6 0.02 72/575 1.5 (1.3–1.8) 3.3 × 10 −3 40/286 1.4 (1.0–1.8) 9.6 × 10 −9 3.0 × 10 −4 1.5 (1.3–1.7) 9.7 × 10−1 LOF + REVEL ≥ 50 2.8 × 10−5 0.068 61/521 1.5 (1.3–1.9) 1.6 × 10−2 34/265 1.3 (1.0–1.7) 2.8 × 10−6 8.7 × 10−2 1.5 (1.3–1.7) 6.6 × 10−1 LOF + REVEL ≥ 75 3.2 × 10 −6 0.014 38/490 1.7 (1.4–2.0) 2.4 × 10 −2 22/243 1.3 (1.0–1.8) 5.7 × 10 −7 1.8 × 10 −2 1.5 (1.3–1.8) 4.2 × 10−1 ABCA1 LOF + REVEL ≥ 75 6.1 × 10 −6 0.019 93/280 1.7 (1.3–2.2) 6.6 × 10 −3 48/159 1.6 (1.1–2.3) 2.6 × 10 −7 8.0 × 10 −3 1.7 (1.4–2.1) 6.3 × 10−1 ADAM10 LOF + REVEL ≥ 50 2.0 × 10 −5 0.051 15/17 3.2 (1.3–8.1) 4.0 × 10 −2 4/4 8.1 (0.6–42.6) 2.8 × 10 −5 8.7 × 10 −1 3.6 (1.5–8.5) 5.5 × 10−1 LOF + REVEL ≥ 75 2.7 × 10−6 0.014 11/12 7.5 (1.4–46.8) 1.5 × 10−1 3/3 5.6 (0.3–41.8) 4.4 × 10−4 1.0 × 100 7.1 (2.6–19.3) 1.1 × 10−1 Listed in this table are the two-sided tests that were significant in stage 1, after multiple testing correction using a Benjamini–Hochberg FDR < 0.1 over 31,204 tests/variant categories. The P values for the burden tests were determined using ordinal logistic regression; a case/control OR was computed for reference. aIn stage 2, we considered only the direction of the AD association observed in stage 1 (that is, one-sided testing). The meta-analysis indicates the combined significance from stages 1 and 2 (data were combined using the fixed-effect inverse variance method); multiple testing correction for the meta-analysis was performed across all 31,204 tests using the Holm–Bonferroni correction (<0.05). Bold text indicates significant P values. Table 2 | GWAS-targeted analysis in a mega-dataset without exome extracts Burden test (variant MAF < 1%) Burden test (variant MAF < 0.1%) Locus sentinel Gene Variant P FDR No. variants/ Case/control P No. variants/ Fraction of Case/control GWAS SNP deleteriousness no. carriers OR (95% CI) no. carriers very rare OR (95% CI) threshold variants, % a SORL1, TREM2, ABCA7 (Table 1 and Supplementary Table 8) SLC24A4/RIN3 RIN3 LOF + REVEL ≥ 25 1.6 × 10−5 0.0003 44/622 1.4 (1.2–1.6) 3.4 × 10−2 42/129 21 1.4 (1.0–2.1) rs7401792 rs12590654 LOF + REVEL ≥ 50 1.0 × 10−5 0.0002 23/583 1.4 (1.2–1.7) 1.5 × 10−2 21/89 15 1.8 (1.2–2.8) a ADAM10, ABCA1 (Table 1 and Supplementary Table 8) PTK2B/CLU CLU LOF + REVEL ≥ 25 5.0 × 10−4 0.005 24/26 3.6 (1.6–8.3) 5.0 × 10−4 24/26 100 3.6 (1.6–8.3) rs73223431 rs11787077 LOF + REVEL ≥ 50 1.1 × 10−3 0.001 14/15 5.4 (1.6–28.6) 1.1 × 10−3 14/15 100 5.3 (1.6–28.6) LOF + REVEL ≥ 75 5.0 × 10 −4 0.005 12/12 9.9 (1.6–44.0) 5.0 × 10 −4 12/12 100 9.8 (1.6–44.0) LOF 2.6 × 10−3 0.02 10/10 7.3 (1.9–27.2) 2.6 × 10−3 10/10 100 7.3 (1.9–27.2) SPDYE3 ZCWPW1 LOF + REVEL ≥ 25 6.1 × 10 −3 0.042 22/77 1.8 (1.2–2.9) 5.0 × 10 −3 21/76 99 1.8 (1.2–2.9) rs7384878 LOF + REVEL ≥ 50 3.1 × 10−3 0.022 16/70 1.9 (1.2–3.1) 3.1 × 10−3 16/70 100 1.9 (1.2–3.1) LOF + REVEL ≥ 75 1.1 × 10 −3 0.001 11/15 5.0 (1.9–13.5) 1.1 × 10 −3 11/15 100 5.0 (1.9–13.5) LOF 7.8 × 10−4 0.008 11/15 5.0 (1.9–13.5) 7.8 × 10−4 11/15 100 5.0 (1.9–13.5) ACE ACE LOF + REVEL ≥ 75 9.0 × 10−4 0.008 38/99 2.0 (1.3–2.9) 9.3 × 10−4 38/99 100 2.0 (1.3–2.9) rs4277405 Genes in all GWAS loci were prioritized as described in the Methods (Supplementary Table 7). Listed are genes for which burden tests were significant in the mega-analysis after multiple testing correction using a Benjamini–Hochberg FDR < 0.05. P values for burden tests were determined using ordinal logistic regression (two-sided tests); a case/control OR was computed for reference. aThese genes also included the SORL1, TREM2, ABCA7, ADAM10 and ABCA1 genes, which were also identified in the rare-variant burden analysis shown in Table 1 and therefore are not shown (see Supplementary Table 8 for the full analysis). Bold text: result of burden test MAF < 0.1% unchanged compared to the burden test MAF < 1%. and also with negligible P value inflation (𝝀 = 1.016; Extended Data presumably because prioritized variants in this gene are extremely few Fig. 3). The effect was tested in the direction observed in stage 1 (one-sided and rare, such that the signal can be confirmed only in larger datasets. test). All genes selected in stage 1 reached P < 0.05 (Table 1, stage 2). The Strikingly, most of these genes also map to GWAS loci (SORL1, stage 2 effect sizes of these genes correlated with those observed in TREM2, ABCA7, ABCA1 and ADAM10). This led us to perform a focused stage 1 (Pearson’s r on log odds = 0.91). We then meta-analyzed stage analysis on GWAS loci, aiming to identify potential driver genes. To 1 + stage 2 across the 13 tests using a fixed-effect inverse variance method maximize statistical power, we merged the full exomes from the stage 1 and corrected for the 31,204 tests performed in stage 1 (Holm–Bonfer- and stage 2 samples into one mega-sample, again with negligible roni) (Table 1). This confirmed the AD association of rare damaging P value inflation (𝝀 = 1.025; Extended Data Fig. 4). We interrogated variants in the SORL1, TREM2, ABCA7, ATP8B4 and ABCA1 genes. The genes that were previously prioritized to drive the AD association in association signal of the ADAM10 gene was not significant exome-wide, the 75 loci identified in the most recent GWAS2 (Supplementary Table 7 Nature Genetics | Volume 54 | December 2022 | 1786–1794 1788 Letter https://doi.org/10.1038/s41588-022-01208-7 a Carrier frequency by age b OR by age 7% Age at onset ≤65 32.0 Age at onset 65–75 Age at onset 75–85 6% Age at onset >85 16.0 Controls 5% OR (95% CI, multinomial test) 8.0 Carrier frequency (95% CI) 4% 4.0 3% 2.0 2% 1.0 Age at onset ≤65 Age at onset 65–75 1% 0.5 Age at onset 75–85 Age at onset >85 Only LOF variants 0% 0.25 1 2 B4 A1 10 3 LU 1 1 2 B4 A1 3 LU 1 E 10 E A7 A7 RL PW RL PW EM EM AC AC N N C C AM AM C C C C P8 P8 RI RI SO SO AB AB W W AB AB TR TR AT AT AD AD ZC ZC c OR by variant frequency d cMAC count by variant frequency 16.0 1,000 *** Singleton Singleton Allele count 2 Allele count 2 22 Allele count 3–5 Allele count 3–5 Allele count 6–10 Allele count 6–10 8.0 Allele count 10+ 800 Allele count 10+ Cases 7 Controls Ordinal OR (95% CI) 4.0 600 3 cMAC 2.0 400 3 190 53 1.0 200 151 19 67 6 22 80 2 27 60 16 6 5 18 9 4 7 13 22 22 1 13 4 3 2 16 14 4 6 2 2 1 2 0 2 8 1 2 00 000 00 0.5 0 1 2 A7 B4 A1 10 3 LU 1 1 2 10 3 LU 1 E B4 A1 E A7 RL PW RL PW EM M AC AC N N C C AM AM C C C C P8 P8 RI RI E SO SO AB AB W W AB AB TR TR AT AT AD AD ZC ZC Fig. 2 | Characterization of gene-specific variant features based on the correction; Supplementary Table 9). c, ORs according to variant frequency. mega-sample. For all variant features, we considered the deleteriousness The rareness of variants in SORL1 was significantly associated with the effect threshold that provides the most evidence for AD association in the size (Supplementary Table 11). d, cMAC by variant frequency: the stacked total meta-analysis. Variant features were investigated in a merged mega-sample number of cases (dark) and controls (light) that carry gene variants with allele (n = 31,905) instead of the meta-sample because this allows for increased frequencies as observed in the mega-sample. The numbers above the bars accuracy for estimations of variant effect sizes for each variant category/ indicate the number of contributing variants. Whiskers: 95% CI. Genes in black: age-at-onset bin (Table 3, refined burden). a, Carrier frequency according to age genes identified to significantly associate with AD in the meta-analysis; gray: at onset. A carrier carries at least one damaging variant in the considered gene. genes not significantly associated with AD in the meta-analysis; blue: genes b, ORs according to age at onset. The effect size significantly decreased with age identified by the targeted GWAS analysis, these were not significantly associated at onset for SORL1, TREM2, ABCA7, ABCA1 and ADAM10 (after multiple testing with AD in the meta-analysis. and Methods). In 67 genes, we observed sufficient prioritized variants the AD associations in these genes persisted when focusing on the (cMAC ≥ 10) to test the burden signal in at least 1 deleteriousness cate- burden of only the very rare variants (MAF < 0.1%), suggesting that the gory (a total of 187 tests). In addition to the genes mentioned above, our rare-variant burden is not in linkage with, and thus independent from, analysis indicated a suggestive signal of increased AD risk in RIN3, CLU, the GWAS sentinel variant. ZCWPW1 and ACE (FDR < 0.05) (Table 2 and Supplementary Table 8); Together, the newly associated genes provide additional evidence these signals will have to be confirmed in a larger dataset. Nevertheless, for a central role for APP processing, lipid metabolism, amyloid-β (Aβ) Nature Genetics | Volume 54 | December 2022 | 1786–1794 1789 Letter https://doi.org/10.1038/s41588-022-01208-7 Table 3 | Mega-analysis: carrier frequency, effect sizes, median age at onset and attributable fraction Mega-analysis Carrier frequency OR (95% CI) Median Attributable age at fraction onset Gene Group No. variant/ EOAD/LOAD/ Case/control EOAD/control LOAD/control (IQR) EOAD/ no. carriers controls, % LOAD, % Primary SORL1 LOF + REVEL ≥ 50 212/418 2.75/1.51/0.68 2.5 (2.0–3.0) 3.3 (2.6–4.1) 2.0 (1.6–2.5) 65 (59–73) 1.91/0.75 analysis —Missense (REVEL 161/354 2.02/1.31/0.66 2.1 (1.7–2.5) 2.5 (2.0–3.2) 1.8 (1.4–2.3) 67 (59–74) 1.22/0.58 50–100) —LOF 51/68 0.78/0.21/0.02 19.8 (11.9–32.7) 40.7 (12.5–133) 11.3 (3.3–38.3) 62 (56–69) 0.76/0.19 TREM2 LOF + REVEL ≥ 25 26/441 2.27/1.90/0.75 2.8 (2.3–3.5) 3.3 (2.6–4.3) 2.6 (2.1–3.3) 69 (62–75) 1.58/1.17 LOF + REVEL ≥ 25 25/404 2.22/1.77/0.62 3.1 (2.6–3.8) 3.8 (2.9–4.9) 2.8 (2.2–3.6) 68 (62–75) 1.63/1.15 (refined) —Missense (REVEL 14/377 2.06/1.63/0.59 3.0 (2.5–3.8) 3.7 (2.8–4.9) 2.7 (2.1–3.6) 68 (62–75) 1.50/1.04 25–100) —LOF 12/66 0.21/0.29/0.16 2.1 (1.2–3.4) 1.7 (0.8–3.5) 2.2 (1.3–3.9) 71 (63–76) 0.09/0.16 —LOF (refined) 11/29 0.16/0.16/0.02 5.6 (2.6–12.1) 5.8 (1.7–19) 5.4 (1.8–16.8) 71 (63–74) 0.13/0.13 ABCA7 LOF + REVEL ≥ 25 351/1,489 6.18/5.04/3.90 1.4 (1.3–1.6) 1.6 (1.4–1.9) 1.3 (1.2–1.5) 69 (61–78) 2.40/1.29 —Missense (REVEL 302/1,372 5.58/4.65/3.63 1.4 (1.3–1.6) 1.6 (1.4–1.8) 1.3 (1.2–1.5) 69 (62–78) 2.06/1.18 25–100) —LOF 49/119 0.62/0.39/0.27 1.7 (1.1–2.4) 2.2 (1.4–3.5) 1.4 (0.9–2.1) 67 (57–74) 0.34/0.11 ATP8B4 LOF + REVEL ≥ 25 94/850 3.56/3.08/2.09 1.4 (1.2–1.6) 1.5 (1.3–1.8) 1.4 (1.2–1.6) 70 (61–78) 1.24/0.84 —Missense (REVEL 74/797 3.35/2.93/1.93 1.5 (1.3–1.7) 1.6 (1.3–1.9) 1.4 (1.2–1.7) 70 (62–78) 1.20/0.84 25–100) —LOF 20/54 0.21/0.16/0.16 1.1 (0.6–1.9) 1.2 (0.6–2.4) 1.0 (0.5–1.8) 70 (59–78) 0.03/−0.01 ABCA1 LOF + REVEL ≥ 75 122/442 1.91/1.50/1.13 1.6 (1.3–2.0) 1.9 (1.5–2.5) 1.5 (1.2–1.9) 70 (60–76) 0.91/0.48 LOF + REVEL ≥ 75 120/282 1.52/1.10/0.52 2.4 (1.9–3.1) 2.9 (2.2–4.0) 2.2 (1.6–2.9) 70 (59–76) 1.01/0.60 (refined) —Missense (REVEL 95/395 1.63/1.32/1.05 1.5 (1.2–1.8) 1.7 (1.3–2.2) 1.4 (1.1–1.8) 70 (61–76) 0.68/0.37 75–100) —Missense (REVEL 93/235 1.24/0.92/0.44 2.3 (1.7–3.0) 2.7 (1.9–3.8) 2.1 (1.5–2.8) 70 (59–76) 0.78/0.48 75–100 (refined)) —LOF 27/47 0.28/0.18/0.08 3.5 (1.9–6.4) 4.7 (2.2–10.3) 2.8 (1.3–6.1) 67 (59–77) 0.22/0.11 ADAM10 LOF + REVEL ≥ 50 19/22 0.23/0.05/0.02 4.7 (2.0–10.8) 9.0 (2.9–28) 2.2 (0.5–8.2) 63 (60–68) 0.20/0.03 GWAS- RIN3 LOF + REVEL ≥ 50 23/583 2.67/2.10/1.62 1.4 (1.2–1.7) 1.6 (1.3–2.0) 1.3 (1.1–1.6) 70 (59–79) 1.04/0.46 targeted analysis —Missense (REVEL 17/577 2.62/2.08/1.61 1.4 (1.2–1.7) 1.6 (1.3–2.0) 1.3 (1.1–1.6) 70 (59–79) 1.01/0.45 50–100) —LOF 6/8 0.06/0.03/0.01 2.1 (0.5–9.3) 2.9 (0.5–18.0) 1.7 (0.3–10.3) 69 (57–86) 0.04/0.01 CLU LOF + REVEL ≥ 25 24/26 0.23/0.09/0.03 3.6 (1.6–8.3) 5.8 (2.0–17.1) 2.5 (0.8–7.6) 63 (58–73) 0.19/0.05 —Missense (REVEL 14/16 0.12/0.06/0.03 2.6 (0.9–7.5) 3.6 (0.9–13.6) 2.1 (0.6–8.0) 68 (58–76) 0.08/0.03 25–100) —LOF 10/10 0.12/0.03/0.01 7.3 (1.9–27.2) 14.2 (2.9–470.4) 3.8 (0.6–122.4) 63 (59–68) 0.11/0.02 ZCWPW1 LOF 11/15 0.15/0.05/0.01 5.0 (1.9–13.5) 9.1 (2.0–42.0) 2.9 (0.8–14.7) 63 (58–81) 0.14/0.03 ACE LOF + REVEL ≥ 75 38/99 0.60/0.39/0.20 2.0 (1.3–2.9) 2.4 (1.5–4.1) 1.7 (1.0–2.7) 67 (60–75) 0.35/0.16 —Missense (REVEL 10/49 0.33/0.22/0.07 3.2 (1.7–5.7) 3.9 (1.8–8.8) 2.7 (1.3–5.9) 66 (61–72) 0.24/0.14 75–100) —LOF 28/50 0.27/0.16/0.14 1.4 (0.8–2.4) 1.7 (0.9–3.4) 1.2 (0.6–2.2) 70 (55–76) 0.11/0.02 For each gene, the AD association statistics are shown for the variant deleteriousness threshold with the most evidence for AD association in the meta-analysis (bold). For genes with sufficient carriers, signals are shown for LOF and missense variants separately (regular text). Individual variants contributing to the burden were validated in a multistage analysis (Supplementary Table 16 and Methods), which resulted in the construction of a refined burden for TREM2 (one variant removed) and ABCA1 (two variants removed). The attributable fraction of a gene is an estimate of the fraction of EOAD and LOAD cases in this sample that have become part of this dataset due to carrying a rare damaging variant in the respective gene (Methods). Note that several variants were excluded from this analysis (that is, due to differential missingness) that would otherwise have been included in the burden. See section 2 of the Supplementary Note for a gene-specific discussion of the variants that contribute to the association with AD and Supplementary Data for the list of variants considered in the burden analysis. Genes shown in bold: the variant burden was significantly associated with AD in the meta-analysis (Holm–Bonferroni <0.05; Table 1). P values for the mega-analysis are shown in Supplementary Table 15. aggregation and neuroinflammatory processes in AD pathophysiol- for the possibility that the missense variants that drive the ATP8B4 ogy. Like ABCA7, ATP8B4 encodes a phospholipid transporter. Rare association do not depend on a LOF effect. ABCA1 also encodes a variants in this gene have been associated with the risk of developing phospholipid transporter; it lipidates apolipoprotein E (APOE)10 and systemic sclerosis, an autoimmune disease9. In the brain, ATP8B4 is poor ABCA1-dependent lipidation of APOE-containing lipoprotein predominantly expressed in microglia. Interestingly, GWAS indicated particles increases Aβ deposition and fibrillogenesis11. In line with a potential association of ATP8B4 with AD2, mainly through the rare this, the rare N1800H LOF variant in ABCA1 was previously associ- missense variant that was most recurrent in our study (G395S). Of note, ated with low plasma levels of APOE and evidence suggested an the OR point estimate for ATP8B4 LOF variants was close to 1, allowing association with increased risk of AD and cerebrovascular disease12. Nature Genetics | Volume 54 | December 2022 | 1786–1794 1790 Letter https://doi.org/10.1038/s41588-022-01208-7 Genome-wide analysis GWAS-targeted analysis LOF 32.0 Missense AD OR (+95% CI) EOAD OR LOAD OR 16.0 8.0 OR (95% CI) 4.0 2.0 1.0 0.5 SORL1 TREM2 ABCA7 ATP8B4 ABCA1 ADAM10 RIN3 CLU ZCWPW1 ACE Fig. 3 | ORs according to age at onset and variant pathogenicity. ORs for LOF significantly larger than the missense burden effect size in the SORL1 and we (red) and missense (yellow) variants as observed in the mega-sample (n = 31,905). observed similar trends in ABCA7 and ABCA1 (Supplementary Table11). Of note, Case/control OR (square, 95% CI), EOAD OR (triangle pointing upward), LOAD for ZCWPW1 only the burden of the LOF variants was significantly associated with OR (triangle pointing downward). Missense variants in the considered gene AD; missense variants are shown for reference purposes (REVEL > 25).Grey: gene appertained to the variant deleteriousness threshold that provides the most was not significantly associated with AD in the meta-analysis. evidence for its AD association (Table 3, refined). The LOF burden effect size was The α-secretase ADAM10 plays a major role in non-amyloidogenic for SORL1, the variants with the lowest variant frequencies had the APP metabolism13. Evidence for the AD association of rare variants in largest effect sizes (Fig. 2c and Supplementary Table 11) and damaging ADAM10 has remained suggestive until now: two rare missense variants variants in ADAM10, CLU and ZCWPW1 were all extremely rare (Fig. 2d). in ADAM10 were reported before to incompletely segregate with LOAD Conversely, we observed that rare but recurrent variants contributed in a few families14 (these variants did not associate with AD in our study; to the AD association of TREM2, ABCA7, ATP8B4 and RIN3 (Fig. 2d). The Supplementary Data) and a nonsense variant in the ADAM10 gene effect sizes of rare coding variant burdens were large compared to the segregated with AD but in a small pedigree15. RIN3 has been associated effect sizes of the GWAS sentinel SNPs (Supplementary Tables 7 and 8). with endosomal dysfunction and APP trafficking/metabolism16,17. CLU Up to 18% EOAD and 14% LOAD cases carried at least 1 predicted damag- (also known as APOJ) affects Aβ aggregation and clearance18 and ACE ing variant in 1 of the 10 genes, compared to 9% of the controls (Sup- is suggested to have a role in Aβ degradation19. Thus far, the role of the plementary Table 12). The fractions of EOAD cases in our sample that histone methylation reader ZCWPW1 is unclear. could be attributed to a rare variant in a specific gene ranged between To better comprehend how these genes associate with AD, we ana- 0.1 and 2.4% (approximately 2%: SORL1, TREM2, ABCA7; approximately lyzed the characteristics of rare damaging variants that contributed to 1%: ATP8B4, ABCA1, RIN3; and <0.5% for the remaining genes); for LOAD the burden using the mega-sample (Fig. 2 and Table 3). For damaging cases, this ranged between 0 and 1.3% (Table 3 and Extended Data Fig. 7). variants in most genes, we observed increased carrier frequencies in We performed an age-matched sensitivity analysis to investigate younger cases and larger effect sizes were associated with an earlier possible effects from other age-related conditions, which supported a age at onset (P = 0.0001) (Supplementary Table 9 and Extended Data role in AD for all ten identified genes (Extended Data Fig. 8). Since APOE Fig. 5). Yet the variants also contributed to an increased risk of LOAD status was used as the selection criterion in several contributing data- (Fig. 2a,b and Table 3). The largest effect sizes were measured for LOF sets, burden tests were not adjusted for APOE-ε4 dosage; in a separate variants in SORL1, ADAM10, CLU and ZCWPW1; carriers of such variants analysis we observed no interaction effects between the rare-variant had the lowest median age at onset, implying a key role for these genes AD association and APOE-ε4 dosage (Supplementary Table 13 in AD etiology (Table 3 and Extended Data Fig. 6). Moderate variant and Methods). Also, the rare-variant burden association was not con- effect sizes were observed for LOF variants in TREM2, ABCA1 and RIN3, founded by somatic mutations due to age-related clonal hematopoiesis while the smallest variant effects were observed in ABCA7, ATP8B4 and (Supplementary Table 14). ACE (Fig. 3 and Table 3). Together, we report ATP8B4 and ABCA1 as new AD risk factors Extremely rare variants contributed more to large effect sizes with exome-wide significance and we report suggestive evidence than less rare variants (P = 0.03; Supplementary Table 10). Indeed, for the association of rare variants in the ADAM10 gene with AD risk. Nature Genetics | Volume 54 | December 2022 | 1786–1794 1791 Letter https://doi.org/10.1038/s41588-022-01208-7 Furthermore, we identified RIN3, CLU, ZCWPW1 and ACE as potential 11. Koldamova, R., Staufenbiel, M. & Lefterov, I. Lack of ABCA1 drivers in GWAS loci, illustrating how analyses of rare protein-modifying considerably decreases brain ApoE level and increases amyloid variants can solve this drawback of GWAS studies20. Larger datasets will deposition in APP23 Mice. J. Biol. Chem. 280, 43224–43235 be required to further confirm these signals. Given the association of (2005). LOF variants with increased AD risk, we suggest that the GWAS risk 12. Nordestgaard, L. T., Tybjaerg-Hansen, A., Nordestgaard, B. G. & alleles in the respective loci might also be associated with reduced Frikke-Schmidt, R. Loss-of-function mutation in ABCA1 and risk activity of the gene, which will have to be evaluated in further experi- of Alzheimer’s disease and cerebrovascular disease. Alzheimers ments. We observed an increased burden of rare damaging genetic vari- Dement. 11, 1430–1438 (2015). ants in individuals with an earlier age at onset. Nevertheless, damaging 13. Saftig, P. & Lichtenthaler, S. F. The alpha secretase ADAM10: variants (including APOE-ε4/ε4) were observed in only 30% of the EOAD a metalloprotease with multiple functions in the brain. Prog. cases (Supplementary Table 12), suggesting that additional damaging Neurobiol. 135, 1–20 (2015). variants are yet to be discovered (Fig. 1b). Further, the effect of struc- 14. Kim, M. et al. Potential late-onset Alzheimer’s disease-associated tural variants such as copy number variants and repetitive sequences mutations in the ADAM10 gene attenuate α-secretase activity. will need to be investigated in future analyses. The associated genes Hum. Mol. Genet. 18, 3987–3996 (2009). strengthen our current understanding of AD pathophysiology. When 15. Agüero, P. et al. α-Secretase nonsense mutation (ADAM10 Tyr167*) treatment options become available in the future, identification of in familial Alzheimer’s disease. Alzheimers Res. Ther. 12, 139 damaging variants in these genes will be of interest to clinical practice. (2020). 16. Shen, R. et al. Upregulation of RIN3 induces endosomal Online content dysfunction in Alzheimer’s disease. Transl. Neurodegener. 9, 26 Any methods, additional references, Nature Research reporting sum- (2020). maries, source data, extended data, supplementary information, 17. Shen, R. & Wu, C. RIN3 binds to BIN1 and CD2AP to increase acknowledgements, peer review information; details of author contri- APP‐CTFS in early endosomes. Alzheimers Dement. 16, e047161 butions and competing interests; and statements of data and code avail- (2020). ability are available at https://doi.org/10.1038/s41588-022-01208-7. 18. Foster, E. M., Dangla-Valls, A., Lovestone, S., Ribe, E. M. & Buckley, N. J. Clusterin in Alzheimer’s disease: mechanisms, genetics, and References lessons from other pathologies. Front. Neurosci. 13, 164 (2019). 1. Gatz, M. et al. Role of genes and environments for explaining 19. Hu, J., Igarashi, A., Kamata, M. & Nakagawa, H. Alzheimer disease. Arch. Gen. Psychiatry 63, 168–174 (2006). Angiotensin-converting enzyme degrades Alzheimer amyloid 2. Bellenguez, C. et al. New insights on the genetic etiology of β-peptide (Aβ); retards Aβ aggregation, deposition, fibril Alzheimer’s and related dementias. Nat. Genet. 54, 412–436 (2022). formation; and inhibits cytotoxicity. J. Biol. Chem. 276, 47863– 3. Holstege, H. et al. Characterization of pathogenic SORL1 genetic 47868 (2001). variants for association with Alzheimer’s disease: a clinical 20. Backman, J. D. et al. Exome sequencing and analysis of 454,787 interpretation strategy. Eur. J. Hum. Genet. 25, 973–981 (2017). UK Biobank participants. Nature 599, 628–634 (2021). 4. Nicolas, G. et al. SORL1 rare variants: a major risk factor for familial early-onset Alzheimer’s disease. Mol. Psychiatry 21, Publisher’s note Springer Nature remains neutral with regard to 831–836 (2016). jurisdictional claims in published maps and institutional affiliations. 5. Cuyvers, E. et al. Mutations in ABCA7 in a Belgian cohort of Alzheimer’s disease patients: a targeted resequencing study. Open Access This article is licensed under a Creative Commons Lancet Neurol. 14, 814–822 (2015). Attribution 4.0 International License, which permits use, sharing, 6. Jonsson, T. et al. Variant of TREM2 associated with the risk of adaptation, distribution and reproduction in any medium or format, Alzheimer’s disease. N. Engl. J. Med. 368, 107–116 (2013). as long as you give appropriate credit to the original author(s) and the 7. Guerreiro, R. et al. TREM2 variants in Alzheimer’s disease. N. Engl. source, provide a link to the Creative Commons license, and indicate J. Med. 368, 117–127 (2013). if changes were made. The images or other third party material in this 8. Bellenguez, C. et al. Contribution to Alzheimer’s disease risk of article are included in the article’s Creative Commons license, unless rare variants in TREM2, SORL1, and ABCA7 in 1779 cases and 1273 indicated otherwise in a credit line to the material. If material is not controls. Neurobiol. Aging 59, 220 e1-220.e9 (2017). included in the article’s Creative Commons license and your intended 9. Gao, L. et al. Identification of rare variants in ATP8B4 as a risk use is not permitted by statutory regulation or exceeds the permitted factor for systemic sclerosis by whole-exome sequencing. use, you will need to obtain permission directly from the copyright Arthritis Rheumatol. 68, 191–200 (2016). holder. To view a copy of this license, visit http://creativecommons. 10. Wahrle, S. E. et al. Overexpression of ABCA1 reduces amyloid org/licenses/by/4.0/. deposition in the PDAPP mouse model of Alzheimer disease. J. Clin. Invest. 118, 671–682 (2008). © The Author(s) 2022 Nature Genetics | Volume 54 | December 2022 | 1786–1794 1792 Letter https://doi.org/10.1038/s41588-022-01208-7 Henne Holstege 1,2,3,4,74 , Marc Hulsman 1,2,3,4,74 , Camille Charbonnier5,74, Benjamin Grenier-Boley6, Olivier Quenez 5, Detelina Grozeva 7, Jeroen G. J. van Rooij8,9, Rebecca Sims 7, Shahzad Ahmad 10,11, Najaf Amin 10,12, Penny J. Norsworthy 13, Oriol Dols-Icardo 14,15, Holger Hummerich13, Amit Kawalia16, Philippe Amouyel 6, Gary W. Beecham17, Claudine Berr 18, Joshua C. Bis 19, Anne Boland20, Paola Bossù 21, Femke Bouwman2,3, Jose Bras22,23, Dominique Campion5, J. Nicholas Cochran 24, Antonio Daniele25, Jean-François Dartigues26, Stéphanie Debette 26,27, Jean-François Deleuze20, Nicola Denning28, Anita L. DeStefano29,30,31, Lindsay A. Farrer 29,31,32,33, Maria Victoria Fernández 34,35,36, Nick C. Fox 37, Daniela Galimberti 38,39, Emmanuelle Genin 40, Johan J. P. Gille41, Yann Le Guen 42, Rita Guerreiro22,23, Jonathan L. Haines 43, Clive Holmes44, M. Arfan Ikram 10, M. Kamran Ikram10, Iris E. Jansen2,3,45, Robert Kraaij 9, Marc Lathrop46, Afina W. Lemstra2,3, Alberto Lleó14,15, Lauren Luckcuck7, Marcel M. A. M. Mannens47, Rachel Marshall7, Eden R. Martin17,48, Carlo Masullo49, Richard Mayeux50,51, Patrizia Mecocci52, Alun Meggy28, Merel O. Mol 8, Kevin Morgan 53, Richard M. Myers24, Benedetta Nacmias 54,55, Adam C. Naj 56,57, Valerio Napolioni 42,58, Florence Pasquier 59, Pau Pastor 60,61, Margaret A. Pericak-Vance 17,48, Rachel Raybould 28, Richard Redon 62, Marcel J. T. Reinders 4, Anne-Claire Richard5, Steffi G. Riedel-Heller 63, Fernando Rivadeneira 9, Stéphane Rousseau5, Natalie S. Ryan37, Salha Saad7, Pascual Sanchez-Juan 15,64, Gerard D. Schellenberg57, Philip Scheltens2,3, Jonathan M. Schott 37, Davide Seripa65, Sudha Seshadri 30,31,66, Daoud Sie41, Erik A. Sistermans41, Sandro Sorbi54,55, Resie van Spaendonk 41, Gianfranco Spalletta 67, Niccolo’ Tesi 1,2,3,4, Betty Tijms2, André G. Uitterlinden 9, Sven J. van der Lee 1,2,3,4, Pieter Jelle Visser2, Michael Wagner68,69, David Wallon70, Li-San Wang57, Aline Zarea70, Jordi Clarimon14,15, John C. van Swieten8, Michael D. Greicius42, Jennifer S. Yokoyama 71, Carlos Cruchaga34,35,36, John Hardy72, Alfredo Ramirez 16,66,68,69,73, Simon Mead 13, Wiesje M. van der Flier 2,3, Cornelia M. van Duijn 10,12, Julie Williams 7, Gaël Nicolas 5,74 , Céline Bellenguez 6,74 & Jean-Charles Lambert 6,74 1 Genomics of Neurodegenerative Diseases and Aging, Human Genetics, Vrije Universiteit Amsterdam, Amsterdam UMC location VUmc, Amsterdam, the Netherlands. 2Alzheimer Center Amsterdam, Neurology, Vrije Universiteit Amsterdam, Amsterdam UMC location VUmc, Amsterdam, the Netherlands. 3 Amsterdam Neuroscience, Neurodegeneration, Amsterdam, the Netherlands. 4Delft Bioinformatics Lab, Delft University of Technology, Delft, the Netherlands. 5Université Rouen Normandie, INSERM U1245 and CHU Rouen, Department of Genetics and CNRMAJ, Rouen, France. 6Université Lille, INSERM, Centre Hospitalier Universitaire Lille, Institut Pasteur de Lille, U1167-RID-AGE facteurs de risque et déterminants moléculaires des maladies liées au vieillissement, Lille, France. 7Medical Research Council Centre for Neuropsychiatric Genetics and Genomics,, Division of Psychological Medicine and Clinical Neuroscience, School of Medicine, Cardiff University, Cardiff, UK. 8Department of Neurology, Erasmus Medical Centre, Rotterdam, the Netherlands. 9Department of Internal Medicine, Erasmus Medical Centre, Rotterdam, the Netherlands. 10Department of Epidemiology, Erasmus Medical Centre, Rotterdam, the Netherlands. 11Leiden Academic Centre for Drug Research, Leiden, the Netherlands. 12Nuffield Department of Population Health Oxford University, Oxford, UK. 13Medical Research Council Prion Unit at University College London, University College London Institute of Prion Diseases, London, UK. 14Department of Neurology, II B Sant Pau, Hospital de la Santa Creu i Sant Pau, Universitat Autònoma de Barcelona, Barcelona, Spain. 15 Biomedical Research Networking Center on Neurodegenerative Diseases, National Institute of Health Carlos III, Madrid, Spain. 16Division of Neurogenetics and Molecular Psychiatry, Department of Psychiatry and Psychotherapy, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany. 17The John P. Hussman Institute for Human Genomics, University of Miami, Miami, FL, USA. 18Université Montpellier, INSERM, Institute for Neurosciences of Montpellier, Montpellier, France. 19Cardiovascular Health Research Unit, Department of Medicine, University of Washington, Seattle, WA, USA. 20Université Paris-Saclay, Commissariat à l’Énergie Atomique et aux Énergies Alternatives, Centre National de Recherche en Génomique Humaine Evry, Gif-sur-Yvette, France. 21Experimental Neuro-psychobiology Laboratory, Department of Clinical and Behavioral Neurology, Istituto di Ricovero e Cura a Carattere Scientifico Santa Lucia Foundation, Rome, Italy. 22Department of Neurodegenerative Science, Van Andel Institute, Grand Rapids, MI, USA. 23Division of Psychiatry and Behavioral Medicine, Michigan State University College of Human Medicine, Grand Rapids, MI, USA. 24 HudsonAlpha Institute for Biotechnology, Huntsville, AL, USA. 25Department of Neuroscience, Catholic University of Sacred Heart, Fondazione Policlinico Universitario A. Gemelli Istituto di Ricovero e Cura a Carattere Scientifico, Rome, Italy. 26Université Bordeaux, INSERM, Bordeaux Population Health Research Center, Bordeaux, France. 27Department of Neurology, Bordeaux University Hospital, Bordeaux, France. 28UKDRI Cardiff, School of Medicine, Cardiff University, Cardiff, UK. 29Department of Biostatistics, Boston University School of Public Health, Boston, MA, USA. 30Framingham Heart Study, Framingham, MA, USA. 31Department of Neurology, Boston University School of Medicine, Boston, MA, USA. 32Department of Epidemiology, Boston University, Boston, MA, USA. 33Department of Medicine (Biomedical Genetics), Boston University, Boston, MA, USA. 34Neurogenomics and Informatics Center, Washington University School of Medicine, St Louis, MO, USA. 35Psychiatry Department, Washington University School of Medicine, St Louis, MO, USA. 36Hope Center for Neurological Disorders, Washington University School of Medicine, St Louis, MO, USA. 37Dementia Research Centre, University College London Queen Square Institute of Neurology, London, UK. 38Fondazione Istituto di Ricovero e Cura a Carattere Scientifico Ca’ Granda, Ospedale Policlinico, Milan, Italy. 39University of Milan, Milan, Italy. 40Université Brest, INSERM, Etablissement Français du Sang, Centre Hospitalier Universitaire Brest, Unité Mixte de Recherche 1078, GGB, Brest, France. 41Genome Diagnostics, Department of Human Genetics, VU University, AmsterdamUMC (location VUmc), Amsterdam, the Netherlands. 42Department of Neurology and Neurological Sciences, Stanford University, Stanford, CA, USA. 43Department of Epidemiology and Biostatistics, Case Western Reserve University, Cleveland, OH, USA. 44Clinical and Experimental Science, Faculty of Medicine, University of Southampton, Southampton, UK. 45Department of Complex Trait Genetics, Center for Neurogenomics and Cognitive Research, Amsterdam Neuroscience, Vrije University, Amsterdam, the Netherlands. 46McGill University and Genome Quebec Innovation Centre, Montreal, Quebec, Canada. 47 Department of Human Genetics, Amsterdam UMC, University of Amsterdam, Amsterdam Reproduction and Development Research Institute, Amsterdam, the Netherlands. 48Dr. John T. Macdonald Foundation Department of Human Genetics, University of Miami, Miami, FL, USA. 49Institute of Nature Genetics | Volume 54 | December 2022 | 1786–1794 1793 Letter https://doi.org/10.1038/s41588-022-01208-7 Neurology, Catholic University of the Sacred Heart, Rome, Italy. 50Taub Institute on Alzheimer’s Disease and the Aging Brain, Department of Neurology, Columbia University, New York, NY, USA. 51Gertrude H. Sergievsky Center, Columbia University, New York, NY, USA. 52Institute of Gerontology and Geriatrics, Department of Medicine and Surgery, University of Perugia, Perugia, Italy. 53Human Genetics, School of Life Sciences, University of Nottingham, Nottingham, UK. 54Department of Neuroscience, Psychology, Drug Research and Child Health University of Florence, Florence, Italy. 55IRCCS Fondazione Don Carlo Gnocchi, Florence, Italy. 56Penn Neurodegeneration Genomics Center, Department of Biostatistics, Epidemiology, and Informatics, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA. 57Penn Neurodegeneration Genomics Center, Department of Pathology and Laboratory Medicine, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USA. 58Genomic and Molecular Epidemiology Laboratory, School of Biosciences and Veterinary Medicine, University of Camerino, Camerino, Italy. 59Université Lille, INSERM, Centre Hospitalier Universitaire Lille, UMR1172, Resources and Research Memory Center (MRRC) of Distalz, Licend, Lille, France. 60Fundació Docència i Recerca MútuaTerrassa and Movement Disorders Unit, Department of Neurology, University Hospital MútuaTerrassa, Barcelona, Spain. 61Memory Disorders Unit, Department of Neurology, Hospital Universitari Mutua de Terrassa, Barcelona, Spain. 62Université de Nantes, Centre Hospitalier Universitaire Nantes, Centre National de la Recherche Scientifique, INSERM, l’institut du Thorax, Nantes, France. 63Institute of Social Medicine, Occupational Health and Public Health, University of Leipzig, Leipzig, Germany. 64Neurology Service, Marqués de Valdecilla University Hospital (University of Cantabria and IDIVAL), Santander, Spain. 65Laboratory for Advanced Hematological Diagnostics, Department of Hematology and Stem Cell Transplant, Lecce, Italy. 66Department of Psychiatry and Glenn Biggs Institute for Alzheimer’s and Neurodegenerative Diseases, San Antonio, TX, USA. 67Laboratory of Neuropsychiatry, Department of Clinical and Behavioral Neurology, Istituto di Ricovero e Cura a Carattere Scientifico Santa Lucia Foundation, Rome, Italy. 68Department of Neurodegenerative Diseases and Geriatric Psychiatry, University Hospital Bonn, Medical Faculty, Bonn, Germany. 69German Center for Neurodegenerative Diseases, Bonn, Germany. 70 Université Rouen Normandie, INSERM U1245 and CHU Rouen, Department of Neurology and CNRMAJ, Rouen, France. 71Memory and Aging Center, Department of Neurology, University of California, San Francisco, CA, USA. 72Reta Lila Weston Research Laboratories, Department of Molecular Neuroscience, University College London Institute of Neurology, London, UK. 73Cluster of Excellence Cellular Stress Responses in Aging-Associated Diseases, University of Cologne, Cologne, Germany. 74These authors contributed equally: Henne Holstege, Marc Hulsman, Camille Charbonnier, Gaël Nicolas, Céline Bellenguez, Jean-Charles Lambert. e-mail: h.holstege@amsterdamumc.nl; m.hulsman1@amsterdamumc.nl; gaelnicolas@hotmail.com; jean-charles.lambert@pasteur-lille.fr Nature Genetics | Volume 54 | December 2022 | 1786–1794 1794 Letter https://doi.org/10.1038/s41588-022-01208-7 Methods donor or frameshifts). Missense and LOF variants were required to have In-depth descriptions of all methods are described in Methods section a ‘moderate’ and ‘high’ variant effect predictor impact classification, of the Supplementary Note. respectively. Then, missense variants were prioritized using REVEL26, annotation obtained from dbNSFP4.1a27 and LOF variants were pri- Sample processing, genotype calling and QC oritized using LOFTEE v.1.0.2 (ref. 28). For the analysis, we considered We collected the exome, whole genome sequencing (WGS) or exome only missense variants with a REVEL score ≥ 25 (score range 0–100) extract sequencing data of a total of 52,361 individuals, brought and LOF variants were annotated as ‘high confidence’ by LOFTEE. Vari- together by the Alzheimer Disease European Sequencing (ADES) con- ants were required to have at least 1 carrier (that is, at least 1 sample sortium, the Alzheimer’s Disease Sequencing Project (ADSP)21 and with a posterior dosage >0.5) and an MAF < 1%, both in the considered several independent study cohorts (Supplementary Table 1). Exome dataset and the Genome Aggregation Database v.2.1 populations (non- extract samples only contained the raw reads that cover the ten genes neurological set). identified in stage 1. Across all cohorts, AD cases were defined accord- ing to National Institute on Aging-Alzheimer’s Association criteria22 for Gene burden testing possible or probable AD or according to National Institute of Neuro- The burden analysis was based on four deleteriousness thresholds logical and Communicative Disorders and Stroke-Alzheimer’s Disease by incrementally including variants from categories with lower and Related Disorders Association criteria23 depending on the date of levels of predicted variant deleteriousness: LOF; LOF + REVEL ≥ 75; diagnosis. When possible, supportive evidence for an AD pathophysi- LOF + REVEL ≥ 50; and LOF + REVEL ≥ 25, respectively. This allowed ological process was sought (including cerebrospinal fluid biomarkers) us to identify the variant threshold providing maximum evidence or the diagnosis was confirmed by neuropathological examination for a differential burden signal. To infer any dependable signal for (Supplementary Table 1). AD cases were annotated with the age at onset a specific deleterious threshold, a minimum of 10 damaging alleles or age at diagnosis (2,014 samples); otherwise, samples were classified appertaining to this deleteriousness threshold was required, that as late-onset AD (366 samples). Controls were not diagnosed with AD. is, a cMAC ≥ 10. Multiple testing correction was performed across All contributing datasets were sequenced using a paired-end Illumina all performed tests (up to four per gene). Burden testing was imple- platform; different exome capture kits were used and a subset of the mented using ordinal logistic regression. This enabled burden testing sample was sequenced using WGS (Supplementary Table 2). to particularly weight EOAD cases since previous findings indicated A uniform pipeline was used to process both the stage 1 and stage that high-impact variants are enriched in early-onset (EOAD) cases 2 datasets. Raw sequencing data from all studies were processed rela- relative to late-onset (LOAD) cases8. This implies that the burden of tive to the GRCh37 reference genome, the read alignments of possible high-impact deleterious genetic variants is ordered according to chimeric origin were filtered and a GATK-based pipeline was used to burdenEOAD > burdenLOAD > burdencontrol. Ordinal logistic regression call variants, while correcting for estimated sample contamination enabled optimal identification of such signals, while also allowing the percentages. Samples were included in the datasets after they passed detection of EOAD-specific burdens (burdenEOAD > burdenLOAD ~ bur- a stringent QC pipeline: samples were removed when they had high dencontrol) and regular case-control signals (burden EOAD ~ burden- missingness, high contamination, a discordant genetic sex annotation, LOAD > burdencontrol). For protective burden signals, the order of the non-European ancestry, high numbers of new variants (with reference signals is reversed, that is, burdenEOAD < burdenLOAD < burdencontrol. to dbSNP v.150), deviating heterozygous/homozygous or transition/ We considered an additive model while correcting for six population transversion ratios. Further, we removed family members up to the covariates, estimated after removal of population outliers. P values third degree and individuals who carried a pathogenic variant in PSEN1, were estimated using a likelihood-ratio test. Genes were selected PSEN2, APP or in other genes causative for Mendelian dementia dis- for confirmation in stage 2 if the FDR for AD association was <0.1 in eases (stage 1-only) or when there was clinical information suggestive stage 1 (Benjamini–Hochberg procedure29). For the GWAS-targeted of non-AD dementia. Variants considered in the analysis also passed analysis, a more stringent threshold was used (FDR < 0.05) due to the a stringent QC pipeline: multiallelic variants were split into biallelic absence of a separate confirmation stage. For the meta-analysis, genes variants; variants that were in complete linkage and near each other were considered significantly associated with AD when the corrected were merged. Further, we removed variants that had indications of an P was <0.05 after family-wise correction using the Holm–Bonferroni oxo-G artifact, were located in short tandem repeat and/or low copy procedure30. Effect sizes (ORs) of the ordinal logistic regression repeat regions, had a discordant balance between reads covering the can be interpreted as weighted averages of the OR being an AD case reference and alternate allele, had a low depth for alternate alleles, devi- versus control and the OR being an early-onset AD case or not. To aid ated significantly from Hardy–Weinberg equilibrium, were considered interpretation, we additionally estimated ‘standard’ case/control false positives based on GATK variant quality score recalibration or ORs across all samples per age category (EOAD versus controls and were estimated to have a batch effect. Variants with >20% genotype LOAD versus controls) and for age-at-onset categories ≤65 (EOAD), missingness (read depth < 6) and differential missingness between 65–70, 70–80 and >80 using multinomial logistic regression, while the EOAD, LOAD and control groups were removed. To account for correcting for 6 PCA covariates. uncertainties resulting from variable read coverage between samples, we analyzed variants according to genotype posterior likelihoods, that GWAS driver gene identification is, the likelihood of being homozygous for the reference allele and For the 75 loci identified in the most recent GWAS2, genes were heterozygous or homozygous for the alternate allele. To account for selected for burden testing based on earlier published gene prioritiza- genotype uncertainty, the burden test was performed multiple times tions. First, gene prioritizations were obtained from Schwarzentruber with independently sampled genotypes and the average P value across et al.31 for 33 known loci. For 28 remaining loci, we obtained the tier these tests is reported. 1 prioritization from Bellenguez et al.2; for loci without prioritiza- tion candidates (14 loci), we selected the nearest gene. In total, 81 Variant prioritization and thresholds protein-coding genes were selected (Supplementary Table 7), of We selected variants in autosomal protein-coding genes that were which 67 genes had sufficient damaging allele carriers to be tested part of the Ensembl basic set of protein-coding transcripts (Gencode for at least 1 variant selection threshold. Gene burden testing was v.19/v.29 (ref. 24); Supplementary Note) and that were annotated by the performed as described above and multiple testing correction to Variant Effect Predictor v.94.542 (ref. 25). Only protein-coding missense identify potential driver genes was performed using the Benjamini– and LOF variants were considered (LOF: nonsense, splice acceptor/ Hochberg procedure, with a cutoff of 5%. Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Validation of variant selection Reporting summary We validated the REVEL variant impact prediction for missense and Further information on research design is available in the Nature the LOFTEE impact prediction for LOF variants for all variants with Research Reporting Summary linked to this article. an MAF < 1%, for which there were at least 15 damaging allele carriers. For protein-modifying variants that were not in the most significant Data availability burden selection of a gene due to a low predicted impact, we investi- The genetic variants analyzed in this study are listed in the Supple- gated whether they, nevertheless, showed a significant AD associa- mentary Data attached to this article. Summary statistics of the stage tion (based on a case/control analysis using logistic regression). Vice 1 analysis are publicly available at Zenodo (https://doi.org/10.5281/ versa, for variants that were in the burden selection, we investigated zenodo.6818051)34 and they can also be downloaded from https://hol- whether their effect size was significantly reduced or oppositely stegelab.eu/data/. For all tests with a cMAC ≥10, this includes Ensembl directed from other missense or LOF variants in the burden selec- gene ID, gene name, variant category, cMAC, P value, beta and s.e.m. tion (Fisher’s exact test). Individual variant effects were analyzed in The ADSP dataset, which includes the ADNI dataset used in this analysis, the stage 1 dataset, followed by a confirmation analysis in the stage is publicly available on request from https://dss.niagads.org/datasets/. 2 dataset. Multiple testing correction was performed per gene, with The accession numbers of the data used in this analysis are: ADSP an FDR < 0.1 used as the threshold for stage 1 and Holm–Bonferroni DBGap: phs000572.v7.p4 (stage 1); ADSP NIAGADS: https://dss.niagads. (P < 0.05) for stage 2. org/datasets/ng00067-v2/ (stage 2). Source data to Figs. 2 and 3 are published alongside this paper. Descriptive measures A variant carrier was defined as an individual for whom the summed Code availability dosage of all the variants in the considered variant deleteriousness The software and algorithms used in the analysis are described in category is ≥0.5 (see Methods section in the Supplementary Note). the Supplementary Note attached to this Letter. Self-contained Carrier frequencies (CFs) were determined as the number of carriers/ code v.0.1.0 can be accessed at https://github.com/holstegelab/ number of total samples. Attributable fraction for cases in an age group shortread_seq_analysis and Zenodo (https://doi.org/10.5281/ was estimated as the probability of a case with an age at onset in the zenodo.6827458)35. age window i being exposed to a specific gene burden (CFcase,gene , i) , multiplied by an estimate of the attributable fraction among the References exposed for these cases: ( ORgene,i −1 )(with the OR being an approximation 21. Bis, J. C. et al. Whole exome sequencing study identifies novel ORgene,i rare and common Alzheimer’s-associated variants involved in of the relative risk)32,33. For large effect sizes, this estimate approaches immune response and transcriptional regulation. Mol. Psychiatry the difference in carrier frequency between cases and controls: 25, 1859–1875 (2020). (CFcase,gene,i ) − (CFcontrol,gene ). 22. McKhann, G. M. et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Sensitivity analyses Institute on Aging-Alzheimer’s Association workgroups on We determined if the observed effects could be explained by age diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. differences between cases and controls. We constructed an 7, 263–269 (2011). age-matched sample, dividing samples into strata based on age/age 23. McKhann, G. et al. Clinical diagnosis of Alzheimer’s disease: at onset, with each stratum covering 2.5 years. Case/control ratios report of the NINCDS-ADRDA Work Group under the auspices in all strata were kept between 0.1 and 10 by downsampling controls of Department of Health and Human Services Task Force on or cases, respectively. Subsequently, samples were weighted using Alzheimer’s Disease. Neurology 34, 939–944 (1984). the ‘propensity weighting within strata method’ (Supplementary 24. Frankish, A. et al. GENCODE reference annotation for the human Note). Finally, a case-control logistic regression was performed both and mouse genomes. Nucleic Acids Res. 47, D766–D773 on the unweighted and weighted case-control labels and estimated (2019). ORs and confidence intervals (CIs) were compared (Extended Data 25. McLaren, W. et al. The Ensembl Variant Effect Predictor. Genome Fig. 8) Also, we determined if somatic mutations due to age-related Biol. 17, 122 (2016). clonal hematopoiesis could have confounded the results. We cal- 26. Ioannidis, N. M. et al. REVEL: an ensemble method for predicting culated for all heterozygous calls in the burden selection the balance the pathogenicity of rare missense variants. Am. J. Hum. Genet. between reference and alternate reads and compared these to refer- 99, 877–885 (2016). ence values (Supplementary Table 14). While APOE was not included 27. Liu, X., Li, C., Mou, C., Dong, Y. & Tu, Y. dbNSFP v4: a as a confounder, we performed a separate APOE interaction analysis comprehensive database of transcript-specific functional (Supplementary Table 13) through a likelihood-ratio test between predictions and annotations for human nonsynonymous and a model label ∼ gene_burden_score + APOE_e4_dosageand an interaction splice-site SNVs. Genome Med. 12, 103 (2020). model label ∼ gene_burden_score + APOE_e4_dosage + APOE_e4_dosage 28. Karczewski, K. J. et al. The mutational constraint spectrum ×gene_burden_score . This test was performed on a reduced dataset, quantified from variation in 141,456 humans. Nature 581, from which datasets in which APOE status was used as the selection 434–443 (2020). criterion were removed. 29. Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Power analysis Soc. Series B Stat. Methodol. 57, 289–300 (1995). Power calculations were performed for ordinal and Firth logistic regres- 30. Holm, S. A simple sequentially rejective multiple test procedure. sion (case-control and EOAD versus rest; Fig. 1b and Supplementary Scand. J. Stat. 6, 65–70 (1979). Table 6). Given the ORs for the EOAD and LOAD cases, and the cMAC per 31. Schwartzentruber, J. et al. Genome-wide meta-analysis, gene, we sampled the number of alleles in the EOAD cases, LOAD cases fine-mapping and integrative prioritization implicate new and controls according to a multinomial distribution. We randomized Alzheimer’s disease risk genes. Nat. Genet. 53, 392–402 these allele carriers across the dataset and performed the burden test (2021). as described above. The power for genes with a cMAC < 10 was set to 0 32. Cole, P. & MacMahon, B. Attributable risk percent in case-control since these genes were not analyzed. studies. Br. J. Prev. Soc. Med. 25, 242–244 (1971). Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 33. LaMorte, W.W. in Measures of Association (Boston University I.E.J., W.M.v.d.F., P.S. and H.Holstege represented the ADC-Amsterdam School of Public Health, 2018). https://sphweb.bumc.bu.edu/ cohort and contributed to sample collection. A.K., S.G.R.-H., M.W. and otlt/mph-modules/ep/ep713_association/EP713_Association8. A.R. represented the AgeCoDe-UKBonn cohort and contributed to html sample collection. C.Charbonnier, O.Q., D.W., A.Z., D.C., A.-C.R., S.R., 34. Holstege, H. et al. Summary statistics for “Exome sequencing G.N., A.B., J.-F.Deleuze, M.L., F.P., E.G., J.-F.Dartigues, R.R., S.D., B.G.-B., identifies rare damaging variants in ATP8B4 and ABCA1 as risk C.Berr, C.Bellenguez, J.-C.L. and P.A. represented the ADES-France factors for Alzheimer’s Disease”. Zenodo (2022) https://doi.org/ cohort and contributed to sample collection. A.C.N., L.A.F., J.L.H., 10.5281/zenodo.6818051 R.Mayeux, M.A.P.-V., J.C.B., L.-S.W., G.W.B., A.L.D.S., E.R.M., S.Seshadri 35. Hulsman, M. & Holstege, H. Software (v.0.1.0) used in “Exome and G.D.S. represented the ADSP cohort and contributed to sample sequencing identifies rare damaging variants in ATP8B4 and collection. M.H., S.J.v.d.L., E.A.S., D.Sie, J.J.P.G., M.M.A.M.M., R.v.S. and ABCA1 as risk factors for Alzheimer’s Disease”. Zenodo (2022) H.Holstege represented the Amsterdam-UMC cohort and contributed https://doi.org/10.5281/zenodo.6827458 to sample collection. O.D.-I., A.L. and J.C. represented the Barcelona SPIN cohort and contributed to sample collection. N.C.F., J.B., R.G. Acknowledgements and J.H. represented the Control Brain Consortium cohort and We thank all the study participants, their families, the participating contributed to sample collection. M.V.F. and C.Cruchaga represented medical staff, general practitioners, pharmacists and all laboratory the Knight-ADRC cohort and contributed to sample collection. personnel involved in patient diagnosis, blood collection, blood D.Grozeva, R.R., S.Saad, N.D., A.M., R.Marshall, L.L., A.D., B.N., P.B., biobanking, DNA preparation and sequencing. The work in this C.M., C.H., D.Galimberti, D.Seripa, P.M., S.Sorbi, G.S., K.M., P.S.-J., manuscript was carried out on the Cartesius supercomputer, which P.P., R.S. and J.W. represented the PERADES cohort and contributed is embedded in the Dutch national e-infrastructure with the support to sample collection. S.A., N.A., R.K., A.G.U., M.A.I., F.R., M.K.I. and of the SURF Cooperative. Computing hours were granted in 2016, C.M.v.D. represented the Rotterdam and ERF cohorts and contributed 2017, 2018 and 2019 to H. Holstege by the Dutch Research Council to sample collection. Y.L.G., V.N., K.M. and M.D.G. represented the (project name: 100-plus; project nos. 15318 and 17232). This research StEP-AD cohort and contributed to sample collection. H.Hummerich, was conducted using the funding obtained by the following study P.J.N., N.S.R., J.M.S. and S.M. represented the UCL-DRC EOAD cohort. cohorts: ADES-FR, AgeCoDe-UKBonn; Barcelona SPIN; AC-EMC; ERF J.N.C., R.M.M. and J.S.Y. represented the UCSF/NYGC/UAB cohort and and Rotterdam; ADC-Amsterdam; 100-plus study; EMIF-90+; Control contributed to sample collection. Brain Consortium; PERADES; StEP-AD; Knight-ADRC; UCSF/NYGC/ UAB; UCL-DRC EOAD; ADSP. Data used in preparation of this article Competing interests were obtained from the Alzheimer’s Disease Neuroimaging Initiative The authors declare no competing interests. (ADNI) database (https://adni.loni.usc.edu/). The investigators within ADNI are listed as supplementary authors and can be Additional information found in Section 5 of the Supplementary Note. Full consortium Extended data is available for this paper at acknowledgements and funding sources are listed in Section 4 of https://doi.org/10.1038/s41588-022-01208-7. the Supplementary Note. Supplementary information The online version Author contributions contains supplementary material available at H.Holstege, G.N. and J.-C.L. jointly supervised the research. https://doi.org/10.1038/s41588-022-01208-7. H.Holstege, M.H., C.Charbonnier, B.G.-B., O.Q., G.N., C.Bellenguez and J.-C.L. were the core writing and analysis group. H.Holstege, Correspondence and requests for materials should be addressed M.H., C.Charbonnier, B.G.-B., O.Q., D.Grozeva, J.G.J.v.R., R.S., S.A., to Henne Holstege, Marc Hulsman, Gaël Nicolas or Jean-Charles N.A., P.J.N., O.D.-I., H.Hummerich2, A.K., J.C., J.C.v.S., J.H., A.R., Lambert. S.M. W.M.v.d.F., C.M.v.D, J.W., G.N., C.Bellenguez and J.-C.L. were the ADES cohort working group. M.H., S.J.v.d.L., M.J.T.R., N.T. and Peer review information Nature Genetics thanks James Lupski and H.Holstege represented the 100-plus study and Netherlands Brain Mike Nalls for their contribution to the peer review of this work. Bank cohorts and contributed to sample collection. P.J.V. represented Peer reviewer reports are available. the EMIF-AD-90 study cohort and contributed to sample collection. J.G.J.v.R., M.O.M., J.C.v.S. represented the AC-EMC cohort and Reprints and permissions information is available at contributed to sample collection. M.H., S.J.v.d.L., F.B., B.T., A.W.L., www.nature.com/reprints. Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 1 | Age, gender, APOE genotype distribution. Age, gender and APOE genotype distribution of all samples, stratified by case/control status. Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 2 | PCA: Sample population compared to 1,000 G reference (see Supplementary Note section 1.3.4). Samples in red are considered population samples. Sample population compared to 1,000 G population population outliers. Samples with only exome-extracts were not included in this samples. First two PCA components of the study samples used for the Stage analysis. 1 and Stage 2 analysis, shown in context of the 1000 Genomes samples for Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 3 | P value inflation in Stage-2 analysis. P value inflation per gene). While not used in this analysis, the threshold for multiple testing in Stage-2 analysis: Quantile-quantile plot for Stage-2 (without exome-extract correction based on FDR < 0.1 is shown for reference. The genomic p-value samples), based on a ordinal logistic burden test (see Methods). Results are inflation was 1.016. Note that causative mutations were not separately removed in shown for all burden tests (n = 20,681) for which at least 10 damaging alleles were Stage-2, as we focused on a specific set of genes. present in this dataset (based on 4 different variant deleteriousness thresholds Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 4 | P value inflation in the mega-analysis dataset. (based on 4 different variant deleteriousness thresholds per gene). For reference, P value inflation in the mega-analysis dataset: Quantile-quantile plot for the the threshold for multiple testing correction based on a false discovery mega-analysis dataset (without exome-extract samples) based on a ordinal rate threshold of 0.1 is shown. P values for the mega-analysis are shown in logistic burden test (see Methods). Results are shown for all burden tests Supplementary Table 15. The genomic p-value inflation was 1.025. (n = 37,710) for which at least 10 damaging alleles were present in this dataset Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 5 | Variant carrier frequency in controls by age last seen. RIN3, CLU, ZCWPW1, ACE (n = 29,727 unique individuals; that is without Variant carrier frequency in controls by age last seen: Carrier frequency in exome-extracts) (Table 3, refined). Black: genes significant in the meta-analysis. controls by age last seen for the variant selection threshold with the strongest Grey: genes not significant in meta-analysis. Blue: genes detected in the GWAS association, as observed in the mega-analysis (n = 31,905 unique individuals); targeted analysis. Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 6 | Age-at-onset by variant deleteriousness category. variant deleteriousness categories with <10 samples are shown individually. The Age-at-onset by variant deleteriousness category: Age-at-onset (median and median age at onset and IQR for the complete mega-analysis dataset is shown on IQR) in the mega-analysis (n = 31,905 unique individuals); RIN3, CLU, ZCWPW1, the right. Black: genes significant in the meta-analysis. Grey: genes not significant ACE (n = 29,727 unique individuals; that is without exome-extracts). Samples in in meta-analysis. Blue: genes detected in the GWAS targeted analysis. Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 7 | Attributable fraction per gene and age-at-onset is an estimate of the fraction of AD cases in a specific age group that have become category. Attributable fraction per gene and age-at-onset category: part of this dataset due to carrying a rare damaging variant in the respective gene Attributable fractions as derived based on the mega-analysis in the mega-analysis (Methods). This estimate accounts only for variants in the burden selection. (n = 31,905 unique individuals); RIN3, CLU, ZCWPW1, ACE (n = 29,727 unique Black: genes significant in the meta-analysis. Grey: genes not significant in individuals; that is without exome-extracts). The attributable fraction of a gene meta-analysis. Blue: genes detected in the GWAS targeted analysis. Nature Genetics Letter https://doi.org/10.1038/s41588-022-01208-7 Extended Data Fig. 8 | Sensitivity Analysis: AD vs Age association. AD vs age-matched and a non-age-matched analysis. Age-matching was performed as Age association: Sensitivity analysis of the gene burden tests (for the most described in the methods. Based on the confidence intervals, we cannot exclude significant deleteriousness thresholds, Table 2) for the mega-analysis dataset that the signals in ACE, ADAM10 and ZCWPW1 are affected by other age-related (RIN3, CLU, ZCWPW1, ACE: without exome-extracts) (respectively n = 31,905 and conditions. Note however, that the signals in ADAM10 and ZCWPW1 are based on n = 29,727 unique individuals). Comparison of the case/control odds ratio of an very few variants, such that confidence intervals are expected to be wide. Nature Genetics